- Client
- Advisory firm
- The work
- Assessment to booked call
- What we built
- Messaging with human approval
- Status
- Documented system
450+ messages handled semi-autonomously each month, saving more than 30 hours.
Message volume from the system's logs, averaged over recent months. The time saving is estimated from handling time per message. Human approval is kept on every reply.
The briefTurn assessment sign-ups into booked discovery calls without anyone chasing by hand.
People who completed the client's assessment, and the replies they sent afterwards, were the starting point for each conversation. The aim was to carry them through to a booked discovery call.
The system had to bring that context into each follow-up while leaving a person in charge of what was actually sent.
What we builtAn approved outbox between the model and the sales system.
We built a reply loop that drafts a response from the assessment result. Incoming messages arrive at a separate public service. Drafts pass through a review queue before they enter an outbox, and the private sales system collects approved replies from there and carries the conversation into its call pipeline.
From assessment to an approved reply
The system in words
Assessment results and incoming messages provide the conversation context. The public receiver supplies messages to a model that drafts a reply. A human approval queue is the default gate before a draft enters the approved outbox. The private sales system polls that outbox for onward handling in the call pipeline. The model drafts; the review queue controls entry to the outbox.
- Assessment and messages
The assessment result and incoming messages give each conversation its context.
- Reply draft
A model drafts the reply using the assessment result. It proposes the wording.
- Human approval
Every draft goes to an approval queue first. Approval by a person is the default.
- Approved outbox
Only approved replies enter the outbox.
- Private sales system
The private sales system polls the outbox for approved replies.
- Call pipeline
The sales system carries the conversation into its call pipeline.
A person approves the reply before it leaves.
This is a semi-autonomous workflow with a human review point. The model writes the draft; a person decides whether it is sent.
- The model drafts each reply within a defined conversation, using the assessment result.
- The approval queue keeps the model's proposal apart from everything that happens next.
- The public service that receives messages and the private sales system have separate roles. They are connected through the approved outbox.
Automation that takes the repeat work, with people keeping the decisions.
If your team spends its day on replies and follow-ups that follow a pattern, we can automate the drafting and keep a person on the approval.
